DIY: Tweeter Ferrofluid Change - with pre- and post- frequency & distortion changes. And on-line amiability among hobbyists.

For this post, I thought I'd discuss a little DIY procedure you might want to think about if you've had your speakers for awhile - like say a decade or more - and know that the tweeter uses "ferrofluid". Not uncommon for tweeters since the 1970's.

Ferrofluid is a substance with ferromagnetic nanoparticles typically suspended inside an organic solvent. These particles are small enough that they're suspended by Brownian motion so over time, the particles do not settle. In tweeters, the ferrofluid surrounds the voice coil which is embedded in the gap between permanent magnets (do electromagnetic speakers use ferrofluid?).

There are benefits to doing this including improving heat transfer away from the wire coil to handle higher power, dampen vibrations (unwanted resonances), and I've heard it said that the ferrofluid can also help suspend the coil within the gap in a way that's better centered (lower distortion).

Now, depending on the type and quality of the ferrofluid, there is the potential that over time this stuff degrades. Some have said that it can dry out, or the consistency becoming more viscous over the years. Here's an interesting video showing degradation of very poor-quality ferrofluid.

About a year ago, I was able to grab some fresh ferrofluid as I was wondering whether it might be time to try refreshing the stuff inside the tweeters of my Paradigm Reference Signature S8 v.3 front speakers which I've been using for over 12 years now.

The reason I thought about this was based on what I saw when running my DSP room correction measurements over the last while. For example, most recently when I ran the Dirac Live Bass Control correction, here's what the frequency response sweep looked like for those front speakers:

Dirac Live speaker measurement of front left and right channels at the main listening position centered between the 2 speakers.

As you can see, the right channel (yellow) dips down more than the left (purple) by up to 4-5dB mainly around 2-3kHz. Being mechanical devices, loudspeakers are not perfect mirror images of each other, there are always manufacturing tolerances a company allows. In time, things could of course break down (or "break in"!) even if the front speakers were perfectly matched when brand new.

Depending on who you ask, I've seen it said that below 500Hz, one expects both channels to perform <5dB of each other and above 500Hz, there should not be much if any deviation >3dB across the audible frequencies. Despite these variations across the audible spectrum with detailed sweeps, for a good hi-fi system, I would not want to see more than 1dB average loudness difference between speakers when sent the same white/pink noise. So, while DSP correction can easily match the two speakers, it's still nice to optimize the performance before processing.

I was suspicious that this could be a ferrofluid issue because back in 2019, I got a tweeter replacement from Paradigm. That right tweeter which is sagging in the upper-mid happens to be the older tweeter suggesting that something seems to be "aging" compared to the new one.

So, the first thing I did was to take a "before" frequency sweep with the UMIK-1 USB microphone at 1m away, tweeter level to get the baseline before I started fooling around. This way I can compare whether the ferrofluid replacement improved performance. I made sure that all DSP processing was turned off.

WARNING: As with any DIY work, there is a risk of damaging your gear which I cannot be responsible for. Please make sure to do some research (this YouTube video might help) and be mindful of any active warranty affected if you proceed.

For these speakers, the tweeters are easy to remove - just 4 screws holding each in place. Here is that right tweeter removed and ready for an "oil change":

Antistatic silicone mat comes in helpful to keep things clean.

Notice the 3 loosened small Torx screws holding the outer metal plate. Let's take the plate off and have a look at the diaphragm/voice coil unit over the magnet:

We can now gently pull off the triangular dome tweeter. Thankfully, the Paradigm tweeter does not have any glue holding the diaphragm down:

Note the round greyish 1" beryllium diaphragm. Let's have a look at the gap in the magnet where the voice coil normally sits with the old ferrofluid inside - that's what we want to change out:

Here's the ferrofluid I got for this job along with some 99% isopropyl alcohol:

As you can see, I bought 4 syringes (CDN$25 each) - each containing 500μL (0.5mL) of Ferrotec Ferrosound APG L11M from Andrew Drouin selling this here in Canada (he's located in Penticton, BC). With the Canadian dollar where it is currently, it's hard to get this stuff locally so I appreciate that Andrew bought this awhile back and is selling off what he grabbed. Here's Andrew's original post on AudiophileStyle, and I see he has a website with more info. Andrew included the gloves in the package which was a nice touch although I didn't actually need to use them for this tweeter.

The Q-tips and a little bit of 99% isopropyl alcohol is used for cleaning up the fragile voice coil later.

The exact amount of ferrofluid you need depends on the tweeter - diameter, thickness of the gap, depth of the gap. What we want is not to completely fill that gap, but basically coat the portion with the voice coil so as to allow good heat transfer. For the Signature S8 v.3 fronts and Signature C3 center channel tweeters, I needed less than 200μL for each, so one syringe is enough for at least two tweeters.

Andrew conveniently included a number of sheets of light bluish card stock which the syringes are taped to for transport. I cut up the paper into about 1-2cm wide strips and stuck them into the gap, running them around, to soak up the old ferrofluid like so:

Be patient, after soaking up with a few sheets, you'll see that all the fluid is out and the gap is clean when the paper isn't stained with any more of the brown stuff.

I then used some Q-tips dipped in 99% isopropyl alcohol to clean old fluid off the tweeter voice coil:

Be very gentle with this! The voice coil is fragile, sticking out maybe about 1/8". As you clean it, make sure that bits of the cotton fiber doesn't get "caught" in the wire. Compared to what's inside the magnetic gap, there should be only a small amount of ferrofluid hugging on to the coil.

Alright, now it's time to put some fresh ferrofluid into the gap:

Ferrofluid magnetically drawn to the gap.

The ferrofluid literally jumps out from the tip of the syringe and will distribute itself quite evenly as you approach the magnetic gap! Don't hold the tip right against the magnet too long since it can be difficult to estimate how much fluid has already entered the gap. Make sure not to overfill as some of it will then spill over when you insert the voice coil. You can check by pulling the voice coil up after inserting it making sure there is no excess fluid sitting on the front of the magnet. As mentioned above, I used less than 1/2 of the 500μL fluid syringe for each tweeter.

So I then pop the tweeter diaphragm back on top in the same way it was dismantled. For this tweeter, there are three indents that help align and "click" the diaphragm in place.

Screw the metal front plate back on with the 3 Torx screws, and then stick it back in the speaker enclosure. Even if you do this very carefully, the whole procedure should take significantly less than an hour if the tweeter is anything like this Paradigm unit with easy access.

Time to remeasure after the ferrofluid "oil change". Again, the MiniDSP UMIK-1 USB microphone is in the same 1m at tweeter axis position, using the appropriate calibration.

Here's what happened to the frequency response (we'll just look at 200Hz up; bass is reinforced by subs):

Nice! That is exactly what I wanted to see - a "lift" in that upper midrange sag! Looks like my hunch was correct and that there was something wrong with the ferrofluid in that older right tweeter such that it degraded over time.

Given my perfectionistic nature, 🙄 I proceeded to also refresh the ferrofluid in my left speaker and center channel as well.

Since ultimately the job is to make sure the 2 front speakers are well-balanced after the DIY, here's what happened:

Very nice! That's what properly matched R+L speakers should measure like. Clearly, less than 1dB deviation between the two speakers until above 10kHz.

At the start of this article I mentioned how ferrofluid is also supposed to be helpful to reduce distortions. Let's have a peek at the right speaker before and after again:

I see this 1.5kHz 3rd harmonic distortion stays around 1-2% tested up to 90dB SPL.

Indeed the normalization of frequency response is also correlated with a substantial reduction in distortion after replacing the ferrofluid.

There is what looks like an inherent ~1% 3rd harmonic centered around 1.5kHz with this speaker. It's also there on the left speaker. While ideally best not to be there, it's not particularly worrisome. Heck, some might even consider the presence of low-order harmonics "euphonic"; adding subjective tonal "richness" as it has been said. 🤔

[As a side note, distortions like this which would be high for DAC or amplifier measurements isn't unusual with speakers. In the case of these Signature S8's, the manufacturer lists the crossover at 2kHz so maybe that 1.5kHz 3rd harmonic reflects a little bit of strain at the upper end of the midrange driver. 

A cursory peek at the NRC loudspeaker measurements on SoundStage! shows distortions like this with the Perlisten Audio R5t, Focal Aria K2 936, KEF Blade 2, Sonus Faber Olympica III, GoldenEar T66, Triangle Borea BR03, Focus Audio Prestige FP90 BE, among others you'll see over a few minutes of looking.

Considering the clearly non-ruler-flat frequency response and amount of potential distortion from speakers, it's a reminder about just how important speakers (and the room) are as contributors to overall sound quality; discussed a few months back.]

Since there could be some "settling in" that might happen, I proceeded to listen to music for the next few nights. Here are the "final" right-left channel frequency response measurements with speaker grilles back on after about 10 hours of music playback:

Clearly the speaker grilles do change the sound. Notice that the grilles toned down the upper-mid frequencies but accentuated two bumps in the treble >8kHz which will add a bit of brightness. That accentuation is mitigated by sitting a bit off-axis which is why I make sure not to toe-in these too much towards the listener sweet-spot. In practice, we can use DSP to shape the sound back to the intended "house curve" and even out those irregularities (discussed previously, these days I use the Dolby Atmos Music Target Curve).

Since waterfall plots look beautiful 😁, let's have a peek into the time domain, along with the step response:

Nothing scary on the waterfall cumulative spectral decay over 3ms. With average of say 80dB SPL, there's a 30dB Y-axis down to the floor at 50dB SPL - a 30dB decline is about 97% loss of sound pressure. There's some stored energy as expected at the >8kHz peaks due to the grilles. Both speakers measure equivalently.

The tweeter and midrange for this speaker are connected in inverted polarity compared to the woofer in positive polarity. This crossover arrangement is similar to Paradigm's Studio 100 v.3 speakers (as measured in Stereophile). There is essentially perfect overlap between the two speakers suggestive of  excellent crossover matching in each, after more than a decade of use.

Since there has been a change in the sound, I'll run Dirac Live with Bass Control to calibrate again. Here's the Dirac Live graph of the measured front channel frequency response taken at the listener "sweet spot" which was where I initially "saw" that L-R imbalance:

Notice BTW that at the "sweet spot" the treble accentuation due to the grilles has been mitigated. This is because we're now measuring from a bit off-axis between the 2 speakers instead of 1m straight in front.

Being able to take measurements is essential for the serious hi-fi audiophile IMO. That's how I saw that there was a clear problem, and now the measurements provide confirmation that the issue has been resolved with a little DIY. It would be ridiculous for anyone to claim that this kind of repair/maintenance could have been "done by ear" with accuracy! The romantic notion that somehow a loudspeaker maker would primarily design "by ear" without strong reliance on objective measurements is silly.

Although I replaced the ferrofluid in both front speakers and also my center channel, it was really only that right tweeter that showed a significant change in frequency response and lower distortion. Hopefully, it's just that I got a marginal tweeter with maybe inadequate or poor quality ferrofluid that caused this issue over time and other Paradigm speakers from the early 2010's are running without issue.

The new Ferrotech APG L11M ferrofluid, said to have high stability, temperature handling, and tolerance of humidity should maintain the tweeter performance well for many more years.

Thanks for the ferrofluid, Andrew - here again is Andrew's website for those especially in Canada who can take advantage of the free shipping.

--------------------

Time for an attitude readjustment?

On online amiability... [With update.]

To end, I wanted to address how some people write/say things online. While I was preparing for this blog post, I came across this forum thread on AudioScienceReview started by Andrew offering the ferrofluid in early 2024. I don't know why the thread was closed to comments because obviously the note from a "Grand Contributor" (I guess just a guy who sticks around and posts a lot?) stating:

"Math check:

$25 / 0.5ul values the whole 100cc bottle at $5,000,000"

Is clearly wrong. If I could, I would have loved to leave a response instead of letting it standing like that.

Yeah, Andrew made a misplaced decimal typo when he wrote in his post ".500ul of ferro". That's an obvious mistake that any of us can make, easily clarified, and typically understood without need for a snarky response. Let's look at that calculation again:

500uL = 0.5mL; so 100mL/0.5mL = 200 x CDN$25 each = CDN$5,000

For that bottle of fluid which costs him US$350 as he transparently stated, plus shipping, plus taxes (usually 12%), duty on top, he probably paid around CDN$600 for the raw fluid. The Canadian dollar conversion has been the worst in 20 years for awhile now. He's now putting time in to carefully pack it up with the syringes and little extras like the paper, gloves, and shipping included within Canada.

The CDN$25/500uL price is very reasonable given how unfortunately difficult it is for us in the Great White North to get high quality stuff in the open market! More than likely lower quality ferrofluid of unknown provenance like this 900uL packet has an asking price >CDN$60 on Amazon.ca. Forget sourcing from places like eBay.ca because with shipping, the options I've seen for Ferrotec APG L11 are invariably more expensive. As such, Andrew is obviously a small-time "dealer" in ferrofluid, using that term with a tongue-in-cheek intent.

I'm all for fellow audiophile hobbyists doing what they can to share goodies like this for DIY purposes. If Andrew can make a few bucks on the side to pay for his research, import efforts, and time since it could take years to sell the 100mL bottle contents depending on demand, we're obviously not looking at big money being made in his small hobby side gig!

I see that the ASR thread was locked and I believe Andrew never got to respond publicly on the thread nor elsewhere on the site as his account also got banned. (Confirmed with Andrew.)

Guys, it's one thing to criticize obvious Snake Oil-peddling unethical companies and uninformed sycophantic audio reviewers perpetuating myths. But give fellow hobbyists a break! I don't know Andrew personally, but my dealings with him have been nothing less than pleasant and helpful. Let's not treat fellow audiophiles in such off-putting, cynical ways. It's simply not fun, intimidating in a way that's just as repugnant as arrogant "high-end" audio dealers, especially for the newer and younger folks interested in a more objective, honest, and dare I say, decent, version of rational audiophilia.

Not that this is just a problem among audiophiles of course. So often online, there's simply an unfortunate lack of "social capital" among participants these days. Even if we disagree on various matters, as hi-fi hobbyists, I trust we're all working towards shared goals and as much as possible, can show warmth, humor, polite manners, and grace.

Regardless of intellectual and scholarly abilities, wealth, and status, first be gentlemen and gentleladies.

I bought another batch recently for future use from Andrew. I see that the syringes are now individually packaged and the paper has been pre-cut for sucking up old fluid.

*** Update January 31, 2025:

I was contacted by ASR around the misunderstanding from the forum side with Andrew's post. Issue rectified and Andrew's account reactivated should he wish to further post there. That's excellent, guys.

--------------------

Okay folks, off to enjoy the music and to get some other work done over the next few weeks. Plus catch up on Season 2 of Squid Game 2 and Severance. 🙂

Always good to end off with some music. Here's Vienna Teng's sweet "Lullabye for a Stormy Night" from Waking Hour (DR9, 2002); another contender for excellent audiophile female vocal test track:

Kygo x Sasha Alex Sloan's "I'll Wait" from Golden Hour (2020) - if you can, have a listen to the multichannel version with subwoofers turned on of course: